What is prescriptive analytics?

Prescriptive analytics is the most advanced level in the data analysis maturity model, after descriptive analysis (what happened), diagnostic analysis (why it happened) and predictive analytics (what will probably happen). It does not stop at predicting an outcome: it recommends the specific action to take to achieve the desired result, often through mathematical optimization or simulation techniques. That is also why this level remains the least common in companies: on top of a reliable predictive model, it requires an explicit formalization of business goals and constraints, work that is often more organizational than technical. The techniques used are mainly mathematical optimization, to find the best solution given a set of constraints, and simulation, to test how a decision behaves before putting it into practice: tools born in operations research, now applied to everyday operational decisions such as which shift to assign or which price to set at a given moment.

See also: Predictive analytics · The use of historical data and statistical or machine learning models to estimate a future outcome: demand, churn, risk.

The key difference from predictive analytics

A predictive model says "demand for product X next week will likely be Y units." A prescriptive system adds "so order Z units from supplier A by Tuesday, the choice that minimizes total cost given the available warehouse constraint." The forecast becomes an input: the added value lies in translating it into a decision already evaluated against the company's real constraints.

Typical techniques

The core of prescriptive analytics is mathematical optimization, linear or nonlinear programming, which finds the best solution given explicit constraints (warehouse capacity, budget, supplier contracts). Alongside it, simulation, for example Monte Carlo simulation, tests how a decision holds up under different scenarios of uncertainty before acting on it.

An optimization model with continuous variables also returns a number that matters: the shadow price of each constraint. It says how much the objective would improve if that constraint were relaxed by one unit, say one more square meter of warehouse space or one more machine hour, with everything else unchanged and within a certain range. Compared with the cost of that unit, it is the figure a manager needs to judge whether expanding capacity pays off; a constraint with slack has a shadow price of zero, and loosening it earns nothing. Mathematically, shadow prices are Lagrange multipliers, called dual values in linear programming (see the chapter on sensitivity analysis in the Bradley, Hax and Magnanti textbook used at MIT).

The link with Decision Intelligence

Prescriptive analytics is the specific optimization or simulation technique. Decision Intelligence is the broader discipline that combines this technique with human judgment and organizational context to arrive at a decision that is made, not a calculated number. An optimization engine proposes the minimum-cost delivery sequence; whether and how to adopt it is decided by Decision Intelligence, which then measures the effects.

The honest limit

A prescriptive system is only as reliable as the constraints and goals it was given: if the model ignores a real constraint, for example a contractual limit with a supplier, the recommendation can be mathematically optimal but operationally wrong. Formalizing constraints and goals well is almost always the hardest part, harder than the algorithm itself.

Why it matters for your business

The leap from forecast to action is where data analysis generates direct economic value, not just insight to read on a report. A company that stops at forecasting leaves a person to turn it into a decision, within human limits of time and computation; a company that reaches prescriptive analytics automates that step and reserves human judgment for the cases that require it.